
# regression
#event study functions----------------------------------------------------------------------
make_twfe_event_study <- function(df, govt_prop, pop_prop, occ_filter, lab1, lab2){
  var_govt_prop = enquo(govt_prop)
  var_pop_prop = enquo(pop_prop)
  
  df <-
    df %>%
    filter_(occ_filter) %>%
    mutate(
      #outcome variables
      govt_prop = !!var_govt_prop,
      pop_prop = !!var_pop_prop,
      over_prop = govt_prop - pop_prop,
      
      #independent variables    
      ever_treat = ((!is.na(treat))*1),
      distance = case_when(ever_treat == 0 ~ 0,
                           TRUE ~ as.numeric(YEAR) - as.numeric(all)),
      grouped_distance = floor(distance/10))
  
  feols(govt_prop ~ i(grouped_distance, ever_treat, ref = -1) + pop_prop | city + YEAR, cluster = ~city, data = df) %>%
    tidy() %>%
    select(term, estimate, std.error) %>%
    add_row(term = "-1",  estimate = 0, std.error = NA_real_) %>%
    mutate(val = str_remove_all(term, "[:alpha:]|\\_|\\:"),
           lab1 = lab1,
           lab2 = lab2)
}

machine <- 
  read.dta13('./replication_file/_4_data//bias_periods.dta') %>%
  mutate(city = paste0(str_to_lower(city), ', ', str_to_lower(state)))


machine_data <- 
  analysis_df_3 %>%
  inner_join(machine %>% 
               #filter(machine==1 & bias2==1) %>%
               group_by(city) %>%
               summarise(start_year = min(year[machine==1 & bias2==1]),
                         end_year = max(year[machine==1 & bias2==1])) %>%
               left_join(analysis_df_3 %>%
                           mutate(year = all) %>%
                           group_by(city,year) %>%
                           summarise())) %>%
  mutate(machine = ifelse(year>start_year & year<=end_year, 1, 0)) %>%
  filter(year>=1900)


bc_machine_out1 <- make_twfe_event_study(machine_data %>% filter(machine==0), govt_white_x_foreign_born, white_x_foreign_born, 'occ == "blue_collar"', lab1 = "Irish", lab2 = "Non-Machine")
bc_machine_out2 <- make_twfe_event_study(machine_data %>% filter(machine==1), govt_white_x_foreign_born, white_x_foreign_born, 'occ == "blue_collar"', lab1 = "Irish", lab2 = "Machine")

machine_plot <-
  bind_rows(bc_machine_out1, bc_machine_out2) %>%
  mutate(val = as.numeric(as.character(val))) %>%
  mutate(pre_post = (val > -1)*1) %>%
  ggplot(aes(x=val, y = estimate, shape = lab2, color = lab2, group = lab2, alpha = factor(pre_post))) +
  geom_point(position = position_dodge(width = 0.5)) +
  geom_errorbar(aes(ymin = estimate - 1.96*std.error, ymax = estimate + 1.96*std.error), 
                position = position_dodge(width = 0.5),
                width = 0) +
  geom_hline(yintercept = 0, linetype = "dotted") +
  #facet_wrap(lab2~.) +
  theme_bw() +
  scale_colour_grey() +
  xlab("Decades Before/After Civil Service Reform") +
  ylab("Effect of Civil Service Reform on Representation") +
  theme(panel.grid.minor = element_blank(), 
        panel.grid.major.x = element_blank(),
        axis.line.y.left = element_blank(),
        legend.position = "right",
        strip.background = element_blank(),
        #legend.title = element_blank(),
        axis.line = element_line(colour = "black"),
        panel.border = element_blank()) +
  scale_x_continuous(breaks = seq(-4, 4, by = 1), 
                     limits = c(-4.5, 4.5)) +
  scale_alpha_manual(values = c(0.4, 1)) +
  guides(alpha = FALSE) +
  ylim(-0.3, 0.3) +
  guides(color = guide_legend("Dominance"), 
         shape = guide_legend("Dominance")) +
  theme(text=element_text(size=9))

#ggsave('../Apps/Overleaf/merit paper/outputs/event_study_machine.png', plot = machine_plot, width=6, height=3)
ggsave('./replication_file/_5_outputs/figures/figure_a8.png', plot = machine_plot, width=6, height=3)

